referencenet initializing warning ?

[lbwang2006]

Some weights of the model checkpoint were not used when initializing ReferenceNet: ['conv_norm_out.bias, conv_norm_out.weight, conv_out.bias, conv_out.weight, up_blocks.3.attentions.2.proj_out.bias, up_blocks.3.attentions.2.proj_out.weight, up_blocks.3.attentions.2.transformer_blocks.0.attn1.to_k.weight, up_blocks.3.attentions.2.transformer_blocks.0.attn1.to_out.0.bias, up_blocks.3.attentions.2.transformer_blocks.0.attn1.to_out.0.weight, up_blocks.3.attentions.2.transformer_blocks.0.attn1.to_q.weight, up_blocks.3.attentions.2.transformer_blocks.0.attn1.to_v.weight, up_blocks.3.attentions.2.transformer_blocks.0.attn2.to_k.weight, up_blocks.3.attentions.2.transformer_blocks.0.attn2.to_out.0.bias, up_blocks.3.attentions.2.transformer_blocks.0.attn2.to_out.0.weight, up_blocks.3.attentions.2.transformer_blocks.0.attn2.to_q.weight, up_blocks.3.attentions.2.transformer_blocks.0.attn2.to_v.weight, up_blocks.3.attentions.2.transformer_blocks.0.ff.net.0.proj.bias, up_blocks.3.attentions.2.transformer_blocks.0.ff.net.0.proj.weight, up_blocks.3.attentions.2.transformer_blocks.0.ff.net.2.bias, up_blocks.3.attentions.2.transformer_blocks.0.ff.net.2.weight, up_blocks.3.attentions.2.transformer_blocks.0.norm2.bias, up_blocks.3.attentions.2.transformer_blocks.0.norm2.weight, up_blocks.3.attentions.2.transformer_blocks.0.norm3.bias, up_blocks.3.attentions.2.transformer_blocks.0.norm3.weight']

is this correct？the training loss is not decreasing，result：

the pose condition is invalid..

[guoqincode]

1. This is correct.
2. Please train more steps in high resolution.
3. If you find our project helpful, please click a star, thank you.

[lbwang2006]

1. This is correct.
2. Please train more steps in high resolution.
3. If you find our project helpful, please click a star, thank you.

have starred! great work

I trained with 768x512, 64 batchsize, 10000 step result, is this ok? should I continue to train or the 10000 step already bad and tune some setting?

[lbwang2006]

the pose condition looks invalid,

[guoqincode]

During my training, I observed that the model learned the reference features before the pose features, and the pose features were learned later.

[lbwang2006]

finally the first stage training failed , the pose condition looks invalid, the max output of poseguilder is much smaller than latent input, I think the learning rate of poseguider needs a bit higher?